Data Quality Dimensions
Data quality dimensions such as completeness, accuracy, consistency, timeliness, validity, and uniqueness turn vague trust into measurable checks with thresholds.
Definition
Data quality dimensions are the properties used to assess whether data is fit for a purpose. Common dimensions include completeness, accuracy, consistency, timeliness, validity, and uniqueness. Each needs an operational definition, a measurement method, a threshold, and an owner to be meaningful.
Context and scope
Without explicit dimensions, data quality debates become anecdotal. A dashboard can be complete but stale, valid but inaccurate, or unique but inconsistent across systems. Naming the dimension clarifies which promise is failing.
Common misconceptions
- Data quality means zero nulls and zero duplicates.
- A single quality score can summarize fitness for every use case.
Anti-patterns
- Monitoring completeness while ignoring timeliness for a real-time pricing feed.
- Publishing quality scores without thresholds, ownership, or a response plan.
Related concepts
Related examples
Authoritative source
How it works
Each dimension needs a concrete test. Completeness might be the share of orders with a non-null customer ID. Accuracy compares values with a trusted source, such as a payment settlement file. Consistency checks whether the same customer has the same segment in two systems. Timeliness measures the age of the newest record against an expected freshness. Validity checks format and domain rules, such as ISO country codes. Uniqueness checks duplicate keys.
Measurements become useful when attached to a threshold and an owner. A check that runs after every load can block publication or open an incident. A trend line shows gradual decay that a single threshold misses. Segment the metric by source or region, because a global average can hide a badly broken partition.
Choosing dimensions
Start from the decision the data supports. A fraud model needs timeliness and accuracy more than completeness of optional fields. A financial report needs accuracy and consistency, and can tolerate batch latency. A customer list needs uniqueness and validity. Document what “good” means for each use case instead of maximizing every metric.
Trade-offs
More checks improve confidence but add compute, alert noise, and pipeline latency. Exact accuracy checks often require a second source, which may not exist. Thresholds set too tight cause false alarms; too loose and they never catch a real problem. Quality work competes with feature delivery, so prioritize datasets that drive money, compliance, or customer trust.
Example and misconception
An orders table is 99.9 percent complete, so the team reports high quality. Yet 12 percent of rows arrive more than six hours late, breaking same-day fulfillment. Completeness was fine; timeliness was the failing dimension. Adding a freshness check with an owner exposed the real issue.
The misconception is that quality is a single score. Fitness depends on the use case, and different dimensions trade off against each other.